Olivier Elemento

The paper develops a scaffold-constrained generative model for protein design, showing it produces high-quality candidates with better ranking scores than previous methods.

This study maps human natural killer cell development in tonsil using single-cell analysis, showing spatial trajectories and tissue niches that support NK cell differentiation.

This study uses spatial proteomics to reveal immune evasion mechanisms in follicular lymphoma, identifying VISTA and CD14-high follicular dendritic cells as therapeutic targets.

TESSERA is a foundation model for the cancer genome that uses self-supervised learning to predict variant pathogenicity, tumour typing, and treatment response, revealing interpretable biomarkers for chemotherapy selection.

This study identifies unique transcriptional features of Hodgkin lymphoma HRS cells, including immune evasion mechanisms and an unfolded protein response, distinguishing them from other B-cell lymphomas.

EXaCT-2 is a whole exome sequencing platform that detects various cancer-related genomic features with high accuracy and sensitivity, outperforming standard assays in detecting sub-clonal mutations and molecular metrics.

This study shows mutant p53 uses its C-terminal domain to reprogram PARP activity, enabling replication stress adaptation and metastasis in TNBC, and identifies mtp53 as a biomarker for PARP inhibitor and DNA-damaging therapy.

BEACON predicts drug side effects and therapeutic outcomes by integrating biomedical knowledge graphs with contextual language models, achieving high accuracy in drug sensitivity and drug-drug interaction predictions.

This study identifies TMEJ as the main repair pathway for APOBEC3A-induced DSBs and shows Polθ inhibition causes synthetic lethality in cancer cells with APOBEC3A activity.

Neoadjuvant IL-15 with PD-1 blockade expands lymph node clonal diversity and preserves stem-like CD8 T cell features, yielding superior antitumor efficacy, durable memory, and reduced metastatic recurrence.

This paper introduces ppmSeq, a sequencing technology that improves accuracy and efficiency in detecting SNVs by sequencing both DNA strands, showing superior performance in clinical applications like ctDNA detection.

SAHA creates a multimodal, subcellular-resolution reference atlas of healthy human tissues, revealing spatial organization and disease-associated disruptions across multiple organ systems.

This study identifies two transcriptionally distinct CD8 T cell populations in tumor-draining lymph nodes that are shaped by neoadjuvant chemoimmunotherapy and influence tumor and blood T cell environments.

This study develops a pan-cancer PDO platform to preserve tumor heterogeneity and identify therapeutic vulnerabilities, finding that 58% of PDOs respond to Talazoparib and combination screens reveal resistance-overcoming drugs.

The paper identifies clonal inflammatory microglia as a cause of neurodegeneration in LCH and ECD, showing their association with neuronal loss and suggesting CSF1R inhibitors as a therapeutic option.

This study profiles early-stage lung adenocarcinoma using multiplexed imaging mass cytometry to reveal two progression pathways: inflammation-driven and fibrosis-driven, with EMT masking malignant cell expansion.

The paper identifies microglia-specific somatic clones with pathogenic variants in Alzheimer's Disease patients that drive neuro-inflammation through the MAPK pathway.

This study identifies immune suppressive barriers in preinvasive lung cancer progression using single-cell analyses, revealing key cellular and spatial changes that could inform therapeutic strategies.

DDI-GPT predicts drug-drug interactions using a large language model enhanced with knowledge graphs, achieving high accuracy and providing explainable insights into interaction mechanisms.

PAIRWISE predicts effective personalized drug combinations in cancer using deep learning and transcriptome data, achieving high AUROC scores in validation tests.

This paper develops BELA, a deep learning model for ploidy prediction in human blastocysts using time-lapse imaging, achieving an AUC of 0.76 on the Weill Cornell dataset.

This study shows that loss of GPR52 in breast cancer cells increases cell clustering, EMT features, and tumor burden in zebrafish, suggesting GPR52 as a potential therapeutic target.

The paper analyzes whole-genome sequencing data from a pan-cancer cohort to define HRD, showing it extends beyond BRCA1/2 mutations and highlights the benefits of WGS over commercial assays for HRD detection.

This study examines how mutagenesis and extrachromosomal DNA contribute to urothelial cancer evolution, revealing clonal APOBEC3 mutations and chemotherapy-induced mutational bursts within ecDNA.

Positional scanning peptide arrays profile substrate sequence specificities across 303 human serine/threonine kinases, enabling computational kinase-substrate annotation and signaling pathway inference across the human phosphoproteome.

This paper introduces UTAG, a method to automatically identify tissue structures in multiplexed imaging data, showing it can detect higher-order tissue organization and relate it to disease states.

This study shows that APOBEC3G drives mutagenesis and clonal evolution in bladder cancer, contributing to mutational profiles in multiple cancer types.

The paper introduces the GA4GH Phenopacket schema to standardize the exchange of phenotypic data for precision medicine applications.

This paper introduces a weakly-supervised method to predict tumor purity from H&E stained slides, achieving high accuracy and enabling clinical applications.

This study benchmarks 26 variations of polygenic risk score models to assess how different assumptions affect disease prediction accuracy using UK Biobank data.